Fine balance

Gil Avery, the research director of a US controls and instrumentation company, published an article in the ASHRAE Journal in October 1990 entitled ‘Balancing a variable flow water system will ruin the control system’. At the time, variable speed pumps were relatively new in both the UK and US. There was consequently considerable debate about how best to implement them.

The message in the article was that if pressure losses across balancing valves are too large then it becomes impossible to select two-port control valves with any real authority and with it the ability to modulate flows. If you lose the ability to modulate the output of terminal units (such as AHUs, fan coils and active chilled beams), then the energy wasted can far outweigh any savings made by reducing pump speed.

The laws of physics have not changed and this principle is still true. However, it appears to have been overlooked on many projects. This exposes the designer to a significant risk if the system subsequently fails to control properly or cannot be commissioned.

The problem is most likely to occur when large differential pressure control valves (DPCVs) are located on floor branches feeding from a vertical riser. This location for the DPCVs is appealing since it neatly divides a large system into self-contained sections, each with their own flow and pressure protection. This makes it very convenient if floors are completed in stages or are let separately. However, if floor branches are of any significant size, then the performance of modulating control valves is likely to be compromised.

To provide good control, a control valve must be sized in relation to the circuit it controls such that the pressure loss across the valve is of the same order as that around the rest of the circuit. If the valve has too little resistance, the flow in the circuit is dominated by the resistance in the rest of the circuit and throttling the valve will have little effect on flow (until virtually closed).

The relationship between the fully open control valve pressure loss and that in rest of the controlled circuit is expressed quantitatively in terms of the ‘valve authority’. CIBSE Controls Guide H recommends that valves are selected such that their authority under normal design conditions is greater than or equal to 0.5, ie the pressure loss through the valve should be at least 50% of the pressure loss around the entire terminal branch in which it sits.

Therefore, given the system dimensions, the pressure loss required across each two-port valve can be calculated and used as the basis for valve sizing. Figure 1 shows the resulting pressure distribution for a 50 m long heating branch serving multiple fan coil units (FCUs) and with a DPCV to hold the total pressure constant.

Referring to Figure 1, and assuming an average pressure loss of 250 Pa/m in the main flow and return pipes, the total pressure loss in these pipes is approximately 12.5 kPa each, ie a total loss of 25 kPa in flow and return pipes combined. The overall pressure loss through terminal branch pipes, heating coils and double regulating valves is likely to be around 5 kPa. Hence, to comply with CIBSE Guide H, two-port control valves should be sized to achieve a pressure loss of at least 30 kPa giving them an authority of 0.5 or greater relative to the overall pressure available across the entire circuit.

To achieve a two-port valve pressure loss (kv) of 30 kPa is often impossible from the normal range of control valves available from manufacturers. Typical heating flow rates are between 0.012 l/s and 0.035 l/s. For most manufacturers’ valve ranges, the minimum kv value is 0.25. At a flow rate of 0.012 l/s, a 0.25 kv valve would have a pressure loss of 3.0 kPa, and at a flow rate of 0.035 l/s it would have a pressure loss of 25.4 kPa. Hence, for most valves the losses will inevitably be well below the target value of 30 kPa and insufficient to give the valve an acceptable control authority in compliance with CIBSE Guide H.

The problem occurs because as you work back up the system, overall branch pressures are increasing. This excess pressure is taken out by the balancing valve which gradually detracts from the control valve authority.

One solution might be to reduce the pressure losses in the flow and return mains pipes by effectively over-sizing these pipes. However, this obviously increases the cost of the system and may result in large low velocity areas which are difficult to flush and clean properly. Alternatively, higher resistance control valves can be sought (with kv values less than 0.25). Such high resistance valves, however, are likely to have very small clearances and will be prone to blockage if debris should enter the system.

In any case, a system of this type is poor in terms of pump energy. The pump has to be bigger in the first place since it needs to overcome the additional pressure loss through a high resistance two-port valve, plus the pressure loss through the DPCV itself, which is typically an additional 20-30 kPa. The likelihood is that the payback period for the variable speed pumps and DPCVs would be very long indeed and may even be negative if excessive energy is wasted due to poor modulating control. This is true of both heating and chilled water systems.

In recognition of these problems, the latest CIBSE Knowledge Series Guide KS7 ‘Variable Flow Pipework Systems’ sets out two simplified approaches to design. These alternatives essentially use smaller, more localised DPCVs or use reverse return layouts. Both solutions effectively reduce the balancing pressures in the system and therefore make it easier to select the two-port control valves.

For designs incorporating DPCVs, there is a rule that states that the constant pressure controlled by the DPCV should not exceed 1.5 times the pressure loss across the most remote terminal branch downstream of the DPCV. In return for this tighter limit on the total branch pressure, the control valve authority can be permitted to drop to 0.3.

This effectively places a limit on the length of the flow and return mains feeding out to the terminals. Using this rule, a 50 m branch would be unlikely; heating systems in particular would be limited to branch lengths of between 10-30 m. Figure 2, on page 59, shows the resulting pressure distribution for a heating branch that is 30 m long (serving approximately 10 FCUs at 3 m spacings). As shown, to comply with the 1.5 rule, the pressure loss across the two-port valve would need to be approximately 8.6 kPa.

This is a significant improvement relative to 30 kPa but it may still not be possible to select two-port valves for flow rates less than 0.02 l/s (8.6 kPa is the pressure loss through a 0.25 kv valve at a flow rate of 0.02 l/s). The designer can then opt to either further reduce the branch length or to specify constant flow four-port valves for all fan coil units with flows less than 0.02 l/s.

The consequence of designing with localised DPCVs is that more pipework will be required in order to create the sub-branches into which DPCVs can be located. Centralised manifolds and flexible pipe distribution can help mitigate some of this cost by providing a saving in installation time.

The other alternative described in CIBSE KS7 is to use a reverse return layout. Although reverse return layouts should not be thought of as ‘self-balancing’, they do nevertheless have the benefit of minimising balancing pressures, thereby making it easier to size two-port valves. As a result, the same limits on terminal branch length need not apply.

Figure 3 shows the typical pressure distribution through a reverse return system 50 m in length. It can be seen that because pressure losses in flow and return mains are in the same direction, the pressure available across each terminal branch stays roughly the same. This means that pressure losses in these pipes do not have to be taken into account when sizing the two-port valves, and there is therefore no limit on branch length.

Furthermore, because of this pressure stability, the use of DPCVs can usually be avoided completely. If pump speed is controlled to maintain the pressure differential across the first terminal branch constant, then that should ensure that all downstream terminal branches also have enough pressure at any part load condition.

Assuming a fairly even load pattern around the system, ie that all two-port valves will modulate closed roughly together, then this method of pump speed control will also ensure that none of the valves need to close against an excessive pressure. Furthermore, the pump will be smaller (compare the overall pressure losses between Figures 1 and 3) and energy savings will be greater without sacrificing the performance of modulating control valves.

The drawback of reverse return layouts can be the extra pipework involved in creating the return leg. However, if systems are arranged in loops or are fed from split risers, then for many layouts, the additional pipework can be minimal. To offset the cost of additional pipework, savings will be achieved due to smaller pumps and the removal of DPCVs.

As more systems are now being designed with variable speed pumps, there is an increasing onus on the designer to ensure that the obvious energy saving benefits are achieved without any detrimental effects on system performance. The solutions recommended in KS7 are simple enough to be encompassed in pipe system design software enabling automatic checking and selection of two-port valves.